RETIREMENT BY SOCIO-DEMOGRAPHIC FACTORS
ANTIDEPRESSANT MEDICATION BEFORE AND AFTER DISABILITY RETIREMENT (SUB-STUDY III)
Of all diagnosed causes of disability retirement investigated in Sub-study III, 30 per cent were mental disorders, over half of which were related to depression. Within the remaining 70 per cent of somatic causes
0 0.5 1 1.5 2 2.5 3 3.5 4
Semi-professionals Routine non-manual employees
Manual workers HR
Base model Hazardous exposures Physical workload
Computer work Job control Job demands
All variables
musculoskeletal diseases constituted the largest group (results not shown).
Retirement type and diagnosed cause strongly influenced the trajectories of antidepressant medication (Figure 9). Among disability retirees the overall direction and timing of changes in the mean level of medication per three-month period were nevertheless largely similar regardless of the diagnosed cause. Four time frames were identified: I) moderate increase 7.5–1.5 years before retirement, II) substantial increase 1.5–0 years before retirement, III) substantial decrease approximately 0–3 years after retirement and IV) moderate decrease 3–7.5 years after retirement. The patterns were more pronounced among those retiring on the grounds of depression, particularly around the time of retirement. In the case of somatic causes, antidepressant medication was at a considerably lower level, pre-retirement increase was less pronounced and there were no clear changes after retirement.
Medication was at an even lower level among those retiring at the statutory age, with a small increase over time but no changes in this trend around the time of retirement.
Figure 9 Antidepressant medication (mean defined daily dose (DDD)/3-month period and its 95% confidence intervals) over four time frames in relation to disability retirement due to any cause (unweighted N=42,937), depression (16% of all diagnosed causes), other mental disorders (14%) and somatic causes (70%) as well as statutory old-age retirement (unweighted N=19,877).
0 10 20 30 40 50 60 70 80 90
-87–90 -81–84 -75–78 -69–72 -63–66 -57–60 -51–54 -45–48 -39–42 -33–36 -27–30 -21–24 -15–18 -9–12 -3–6 +0–3 +6–9 +12–15 +18–21 +24–27 +30–33 +36–39 +42–45 +48–51 +54–57 +60–63 +66–69 +72–75 +78–81 +84–87
MeanDDD
Months before and after retirement
All causes Depression Other mental Somatic Old-age
Time frame I Time frame II
Time frame III Time frame IV
Results
Socio-demographic factors largely modified the trajectories in antidepressant medication before and after all-cause disability retirement, the variations depending to a great extent on the various diagnosed causes of retirement among the different groups (results not shown). Analyses of the modifying effects of these factors were therefore carried out separately for those retiring due to depression (Table 5) and somatic causes (Table 6).
According to the regression analyses, the observed increases in antidepressant medication before and decreases after disability retirement were all statistically significant in these study populations overall. Moreover, even though no change was observed following retirement on somatic grounds in Figure 9, adjustment for the calendar year in the regression models accounted for an increasing secular trend. This resulted in stronger post-retirement decreases, thereby also disclosing decreases after retirement on somatic grounds (Table 6). This secular trend also fully explained the small increase in antidepressant medication observed in Figure 9 among
Table 5 Change in antidepressant medication (defined daily dose (DDD)/3-month period) over time in relation to disability retirement on the grounds of depression and the modifying effects of the socio-demographic factors.
Time in relation to disability retirement 7.5–1.5 years
Age at retirement p=0.000 p=0.000 p=0.000 p=0.000
30–44 0.99 11.88 -2.10 -0.84
45–54 0.54 10.13 -2.85 -1.42
55–64 0.28 6.94 -3.88 -0.39
Social class p=0.004 p=0.637 p=0.000 p=0.933
Upper non-manual 0.99 9.62 -3.86 -0.97
Lower non-manual 0.48 9.75 -3.70 -1.08
Manual 0.39 9.26 -2.15 -0.99
Entrepreneurs 0.56 8.45 -2.72 -1.29
Living arrangements p=0.065 p=0.093 p=0.758 p=0.001
With partner 0.47 9.85 -2.96 -1.34
Alone 0.76 8.84 -3.13 -0.55
Single parent 0.35 10.11 -3.35 -0.42
Adjusted for the calendar year and mutually adjusted for each socio-demographic factor The p-values are calculated for the interaction between a socio-demographic factor and change in antidepressant medication (bold: p<0.05)
those retiring at the statutory age. Consequently, the regression analyses indicate no changes either before or after statutory retirement, a result that was also consistent across all socio-demographic groups (results not shown).
Age at disability retirement modified the trajectories the most. Increases in antidepressant medication before retirement on the grounds of both depression and somatic causes were much stronger in younger age groups (Tables 5 and 6), among which the decreases were also smaller after depression-based retirement (Table 5). There was even a continuing increase in medication among those under the age of 45 whose retirement was attributable to somatic causes (Table 6). Antidepressant medication was already at a relatively low level among those over the age of 54 three years after retirement (results not shown), which explains the much slower decreases in subsequent years (Tables 5 and 6).
Table 6 Change in antidepressant medication (defined daily dose (DDD)/3-month period) over time in relation to disability retirement on somatic grounds and the
modifying effects of the socio-demographic factors.
Time in relation to disability retirement 7.5–1.5 years
before
1.5–0 years before
0–3 years after
3–7.5 years after
Total 0.04 0.73 -0.09 -0.08
(95% CI) (0.01, 0.06) (0.66, 0.79) (-0.14, -0.04) (-0.13, -0.03)
Gender p=0.167 p=0.406 p=0.061 p=0.682
Men 0.03 0.75 -0.13 -0.08
Women 0.06 0.68 -0.03 -0.10
Age at retirement p=0.001 p=0.000 p=0.000 p=0.120
30–44 0.13 1.08 0.32 0.04
45–54 0.06 0.89 -0.10 -0.15
55–64 0.01 0.51 -0.18 -0.07
Social class p=0.009 p=0.000 p=0.829 p=0.640
Upper non-manual 0.17 0.96 -0.02 -0.09
Lower non-manual 0.03 0.97 -0.10 -0.14
Manual 0.02 0.59 -0.07 -0.08
Entrepreneurs 0.08 0.67 -0.14 -0.01
Living arrangements p=0.919 p=0.075 p=0.259 p=0.434
With partner 0.04 0.68 -0.07 -0.09
Alone 0.03 0.84 -0.18 -0.11
Single parent 0.03 0.99 -0.15 0.12
Adjusted for the calendar year and mutually adjusted for each socio-demographic factor The p-values are calculated for the interaction between a socio-demographic factor and change in antidepressant medication (bold: p<0.05)
Results
The modifying effects of other socio-demographic factors were more modest. Changes in antidepressant medication in the time frames surrounding depression-based disability retirement were slightly stronger among the men (Table 5). There were no gender differences in retirement on somatic grounds (Table 6).
Decrease in antidepressant medication immediately after depression-based disability retirement was somewhat stronger in the higher social classes, among which the increase was also stronger 7.5–1.5 years before retirement (Table 5). The increase in medication before retirement on somatic grounds was also strongest among the higher social classes (Table 6).
The trajectories for entrepreneurs were close to average levels (Tables 5 and 6).
The only modifying effect of living arrangements was that the decrease in antidepressant medication 3–7.5 years after depression-based disability retirement was strongest among people living with a partner (Table 5).
EXCESS MORTALITY AFTER DISABILITY RETIREMENT ON THE GROUNDS OF MENTAL DISORDERS (SUB-STUDY IV)
Almost three per cent of the study population retired on the grounds of mental disorders during the follow-up reported in Sub-study IV (Table 7).
Depression was the more specific diagnosed cause in almost half of the male cases, and almost two thirds of the female cases. The age-adjusted mortality rate was around three-fold among both males and females who retired due to mental disorders compared to the rest of the study population, this excess mortality being more pronounced in the case of mental disorders other than depression. Alcohol-related causes, suicide and other unnatural causes accounted for a large proportion of the deaths among the retirees, whereas circulatory diseases and cancer were more prevalent among those with no mental-health-based retirement history. Suicide and alcohol-related causes of death were relatively common in relation to retirement on the grounds of depression and other mental disorders, respectively.
Excess mortality following disability retirement on the grounds of mental disorders was larger in the case of unnatural and alcohol-related deaths and generally also in the case of mental disorders other than depression (Figures 10 and 11). Excess mortality related to suicide was nevertheless particularly large after depression-based retirement, and especially among women. There was no excess risk of cancer mortality except among men retired on the grounds of mental disorders other than depression. The magnitude of the excess mortality nevertheless varied within the category of other mental disorders, being particularly high among those retired on the grounds of alcohol-related disorders (ICD-10 F10) (results not shown).
Table 8 shows the modifying effects of age, social class and living arrangements on excess mortality from natural causes following retirement on the grounds of depression and other mental disorders among men and
women. Both absolute and relative differences are considered through mortality rates and HRs, respectively. The HRs indicate excess mortality within their own socio-demographic group among those who retired on the grounds of mental disorders compared to those with no such retirement history, and the p-values show the statistical significance of the interactions between each socio-demographic factor and mental-health-based retirement.
The mortality rate increased strongly with increasing age regardless of the retirement status, and was particularly high among the oldest group retired for reasons other than depression. However, excess mortality following retirement on the grounds of mental disorders was higher in the younger age groups. In terms of social class, the mortality rate followed a rather typical
Table 7 Study population (% & unweighted N), mortality (age-adjusted rates per 1000 person years using the whole study population as the standard population and the unweighted number of deaths), and causes of death (%) by gender and mental-health-based disability-retirement status.
Disability retirement due to mental disorders
Yes No
MEN All mental Depression Other mental
% 2.6 1.2 1.4 97.4
N 6,712 2,608 4,104 208,909
Mortality rate 22.2 16.5 29.4 6.6
Number of deaths 3,328 1,014 2,314 73,461
Cause of death (%)
Circulatory 21 21 22 34
Cancer 11 13 10 25
Other natural 11 8 13 13
Alcohol 31 25 34 12
Suicide 10 18 7 6
Other unnatural 15 15 15 10
Total 100 100 100 100
WOMEN
% 2.9 1.8 1.1 97.1
N 5,589 3,202 2,387 171,775
Mortality rate 8.6 7.2 11.1 2.8
Number of deaths 1,489 675 814 33,085
Cause of death (%)
Circulatory 15 14 15 23
Cancer 20 27 14 45
Other natural 16 10 21 17
Alcohol 22 15 28 6
Suicide 16 21 11 3
Other unnatural 12 12 12 6
Total 100 100 100 100
Results
Figure 10 Excess mortality (hazard ratios (HR) and their 95-per-cent confidence intervals) following disability retirement on the grounds of depression and other mental disorders compared to those with no such retirement history (HR=1.00), by cause of death, men.
Figure 11 Excess mortality (hazard ratios (HR) and their 95-per-cent confidence intervals) following disability retirement on the grounds of depression and other mental disorders compared to those with no such retirement history (HR=1.00), by cause of death, women.
Adjusted for age, social class, and living arrangements 0
Adjusted for age, social class, and living arrangements 0
pattern, being higher in the lower classes. However, in the case of women retired on the grounds of depression the rate varied little by social class. In these women excess mortality was higher among those in the higher social classes, whereas it was similar across the classes in the rest of the retired groups. The mortality rates of both men and women living alone were higher regardless of retirement status. There was no variation in excess mortality by living arrangements among the women. Among the men, excess mortality following retirement on the grounds of depression and other mental disorders was largest among those living with a partner and children and smallest among those living only with a partner, although among the latter retired group the interaction effect was not statistically significant.
In the case of unnatural and alcohol-related causes of death, age differences in the mortality rate varied somewhat by retirement status (Table 9). It was typically lowest among those under 45, and this absolute difference from the older age groups was particularly pronounced among men retired on the grounds of mental disorders other than depression. This was largely attributable to the high proportion of alcohol-related disorders as diagnosed causes of disability retirement among older groups (results not shown). The mortality rate among the 45-54-year-olds was typically higher than or similar to the rate among the over-54s, with the exception of women retired on the grounds of depression among whom the rate decreased with increasing age.
There was also variation in excess mortality by age group. Following depression-based retirement it decreased with increasing age among both men and women. In the case of those retired on the grounds of other mental disorders excess mortality was similar in all age groups among the women, but somewhat smaller among the men in the youngest compared to the older groups. Social-class differences in the mortality rate varied according to the retirement status. Typical gradients were found among those with no mental-health-based retirement history, as well as among women retired on the grounds of disorders other than depression: in other words the rate was lowest among the upper non-manual class and highest among the manual class. In the other retired groups, in other words men retired on the grounds of both depression and other mental disorders as well as women retired due to depression, the mortality rates were relatively similar across the classes. In these groups, excess mortality was higher in the higher social classes. The mortality rate was highest among those living alone in all retirement-status categories, and in absolute terms it was particularly high among men retired on the grounds of mental disorders who were living alone. Excess mortality following disability retirement on these grounds was smallest among those living alone and largest among those living with a partner and children, regardless of gender and the type of mental disorder.
Results
Table 8 Mortality rates per 1000 person years and hazard ratios (HR) for mortality from natural causes among men and women, by mental-health-based disability-retirement status, and the modifying effects of age, social class and living arrangements.
P-value for interaction p=0.000 p=0.000 p=0.002 p=0.009
Social class
Upper non-manual Rate 2.6 2.7 6.6 1.5 2.4 3.7
HR 1.00 1.06 2.57 1.00 1.66 2.52
Lower non-manual Rate 3.5 4.8 7.8 1.8 2.3 4.1
HR 1.00 1.36 2.20 1.00 1.30 2.26
Manual Rate 4.6 5.4 10.9 2.4 2.6 5.1
HR 1.00 1.17 2.36 1.00 1.10 2.12
Entrepreneur Rate 3.7 3.6 9.1 2.2 1.7 3.6
HR 1.00 0.98 2.46 1.00 0.77 1.65
P-value for interaction p=0.508 p=0.790 p=0.030 p=0.650
Living arrangements
Partner and children Rate 1.7 2.5 4.8 0.9 1.0 1.6
HR 1.00 1.48 2.87 1.00 1.14 1.79
Partner only Rate 2.1 1.9 4.2 1.1 1.2 2.2
HR 1.00 0.91 1.96 1.00 1.09 2.03
Alone Rate 4.1 5.2 10.0 1.5 2.0 3.2
HR 1.00 1.27 2.46 1.00 1.35 2.15
P-value for interaction p=0.014 p=0.121 p=0.163 p=0.430
Adjusted for age, social class and living arrangements
The p-values are for the interaction between a socio-demographic factor and retirement calculated separately for disability retirement on the grounds of depression and other mental disorders (bold: p<0.05)
Table 9 Mortality rates per 1000 person years and hazard ratios (HR) for mortality from unnatural and alcohol-related causes among men and women by mental-health-based disability-retirement status, and the modifying effects of age, social class and living arrangements.
HR 1.00 5.55 4.21 1.00 10.78 7.84
45–54 Rate 2.2 8.8 11.8 0.4 2.3 4.1
HR 1.00 3.95 5.27 1.00 5.35 9.55
55– Rate 2.4 6.5 12.6 0.3 1.3 2.8
HR 1.00 2.74 5.29 1.00 3.86 8.20
P-value for interaction p=0.000 p=0.027 p=0.002 p=0.319
Social class
Upper non-manual Rate 0.8 5.3 7.6 0.3 2.5 2.7
HR 1.00 6.37 9.14 1.00 9.91 11.00
Lower non-manual Rate 1.2 4.7 7.3 0.4 2.1 3.0
HR 1.00 3.86 5.96 1.00 5.76 8.41
Manual Rate 1.8 6.0 9.0 0.6 2.2 4.9
HR 1.00 3.25 4.88 1.00 3.79 8.33
Entrepreneur Rate 1.3 4.2 6.3 0.4 3.1 3.9
HR 1.00 3.30 4.93 1.00 7.43 9.31
P-value for interaction p=0.000 p=0.000 p=0.000 p=0.790
Living arrangements
Partner and children Rate 0.7 4.5 6.3 0.2 1.8 2.3
HR 1.00 6.81 9.56 1.00 10.74 13.54
Partner only Rate 1.1 4.0 8.2 0.5 2.4 5.7
HR 1.00 3.69 7.61 1.00 5.22 12.48
Alone Rate 3.8 12.4 17.2 0.9 3.8 5.7
HR 1.00 3.29 4.56 1.00 4.50 6.69
P-value for interaction p=0.000 p=0.000 p=0.000 p=0.000
Adjusted for age, social class and living arrangements
The p-values are for the interaction between a socio-demographic factor and retirement calculated separately for disability retirement on the grounds of depression and other mental disorders (bold: p<0.05)
Discussion